Tyrosine Sulfation Is a Trans-Golgi-Specific Protein Modification Patrick A

Home , Sulfation, Tyrosine sulfation

Tyrosine Sulfation Is a trans-Golgi-specific Protein Modification Patrick A. Baeuerle and Wieland B. Huttner Cell Biology Program, European Molecular Biology Laboratory, 6900 Heidelberg, Federal Republic of Germany

Abstract. The trans-Golgi has been recognized as heavy chain detectable by sulfate labeling, as well as having a key role in terminal glycosylation and sorting the heavy chain sulfated in a cell-free system in the of proteins. Here we show that tyrosine sulfation, a absence of vesicle transport, already contained galac- frequent modification of secretory proteins, occurs tose and sialic acid. Third, sulfate-labeled IgM moved specifically in the trans-Golgi. The heavy chain of im- to the cell surface with kinetics identical to those of munoglobulin M (IgM) produced by hybridoma cells galactose-labeled IgM. Lastly, IgM labeled with sul- was found to contain tyrosine sulfate. This finding al- fate at 20~ was not transported to the cell surface at lowed the comparison of the state of sulfation of the 20~ but reached the cell surface at 37~ The data heavy chain with the state of processing of its suggest that within the trans-Golgi, tyrosine sulfation N-linked oligosaccharides. First, the pre-trans-Golgi of IgM occurred at least in part after terminal glycosy- forms of the IgM heavy chain, which lacked galactose lation and therefore appeared to be the last modifica- and sialic acid, were unsulfated, whereas the trans- tion of this constitutively secreted protein before its Golgi form, identified by the presence of galactose and exit from this compartment. Furthermore, the results sialic acid, and the secreted form of the IgM heavy establish the covalent modification of amino acid side chain were sulfated. Second, the earliest form of the chains as a novel function of the trans-Golgi.

major function of the Golgi complex is the covalent expect that the sulfation reaction should occur in one or more modification of proteins destined for lysosomes, of those compartments that are part of the secretory pathway, secretory granules, and the plasma membrane. Gol- i.e., the rough endoplasmic reticulum, the Golgi complex, gi-specific processing of asparagine-linked oligosaccharides and the various transport vesicles and storage granules. In- is particularly well characterized (for review see Kornfeld deed, a recent study indicated that after subcellular fraction- and Kornfeld, 1985; Farquhar, 1985). These reactions in- ation tyrosylprotein sulfotransferase activity is highest in clude (a) the specific removal of carbohydrate units, (b) the Golgi-enriched membrane fractions and that the active site addition of new carbohydrates, and (c) the addition of resi- of this membrane-bound enzyme is oriented toward the Gol- dues other than carbohydrates, such as phosphate to N-linked gi lumen (Lee and Huttner, 1985). oligosaccharides. The detailed characterization of these mod- Among the subcompartments of the Golgi, the trans-most ification reactions has contributed significantly to the under- structures, referred to as the trans-Golgi network (Grifliths standing of the structural and functional organization of the and Simons, 1986), have recently received a great deal of at- Golgi complex, which is currently viewed to comprise at tention. The trans-Golgi network has been proposed to have least three distinct subcompartments, the cis-, medial, and a key role in the sorting of lysosomal, plasma membrane, trans-Golgi (for reviews see Farquhar and Palade, 1981; Tar- and secretory proteins into membrane vesicles that transport takoff, 1982; Dunphy and Rothman, 1985; Farquhar, 1985). them to their specific destination. To understand the molecu- Compared with the information about the localization of pro- lar mechanisms involved in these sorting processes, it would cessing reactions involving N-linked oligosaccharides in the be helpful to identify protein components specific to the Golgi subcompartments, little is known about the localiza- trans-Golgi network. Such protein components would pro- tion of amino acid side chain modifications in the specific vide new markers for the trans-Golgi network that could subcompartments of the Golgi. be used for a more refined description of this compartment Recent work has indicated that the sulfation of tyrosine and that could also be employed for its isolation. Further- residues is a widespread protein modification and that secre- more, the characterization and comparison of several differ- tory proteins are the major physiological substrates for the ent trans-Golgi network-specific proteins may help to iden- sulfating enzyme(s) (Huttner, 1982; Hille et al., 1984; for re- tify domains on these proteins that are involved in their view see Huttner and Baeuerle, 1987). One would therefore specific localization and retention in the trans-Golgi net- Dr. Baeuede's present address is Whitehead Institute, Nine Cambridge Cen- work. Here, we report, using IgM as a model protein, that ter, Cambridge, MA 02142. tyrosine sulfation specifically occurs in the trans-Golgi and

The Rockefeller University Press, 0021-9525/87/12/2655/10 $2.00 The Journal of Cell Biology, Volume 105 (No. 6, Pt. 1), Dec. 1987 2655-2664 2655 apparently is the last modification of this secretory protein a 10% CO2 atmosphere at 370C. Immediately after resuspension and after before its exit from the trans-Golgi. the times indicated, aliquots of the cell culture were removed and boiled in SDS-lysis buffer (see below), lgM was immunoaffinity-purified from the lysates as described in Immunoaffinity Purification of lgM and the radioactiv- Materials and Methods ity in the various forms of the IgM heavy chain determined by liquid scintillation counting after SDS-PAGE under reducing conditions. For pulse-labeling with pH]tyrosine, cells were resuspended at a den- Isotopes sity of 1 • l0 s cells/mi in labeling DME supplemented with 10% dialyzed FCS and allowed to equilibrate for 1 h in the incubator. Cells were then la- L-[3SS]methionine (carrier-free), L-[3,5-3H]tyrosine (1.5 TBq/mmol), L- beled for 10 rnin with 1.5 MBq/ml [3H]tyrosine, cooled on ice, and cen- [~C(U)]tyrosine (1.85 GBq/mmol), [3SS]sulfuric acid or sodium [3SS]sul- trifuged at 4~ for 10 rain at 150 g. The cells were resuspended in chase fate (carrier-free), L-[6-3H]fucose (0.93 TBq/mmol), v-[4,5JH(N)]galac - medium (complete DME, supplemented with 10% FCS) that had been pre- tose (1.7 TBq/mmol), and D-[2-3H(N)] mannose (0.89 TBq/mmol) were equilibrated in a 10% CO2 atmosphere at 37~ and distributed onto six obtained from New England Nuclear (Dreieich, FRG). 3'-Phosphoadeno- dishes that were immediately returned to the incubator. After the indicated sine 5'-phospho-[35S]sulfate (PAPS; 1 4.81 TBq/mmol) was kindly provided periods of chase, dishes were placed on ice. Cells and media were further by Christof Niehrs of this institution. processed as described in Immunoaffinity Purification of IgM. For pulse double-labeling, cells were resuspended at a density of 1 • Cell Culture 10s cells/ml in labeling DME supplemented with 0.1% dialyzed FCS and allowed to equilibrate for 1 h. Cells were then labeled with a mixture of 18.5 The hybridoma cell line studied is the clone 81-O4 of Sommer and Schach- MBq/ml [35S]sulfate and 12.3 MBq/ml [3H]galactose for 5 min in the incu- ner (1981), which secretes lgM directed against the cell surface antigen 04. bator. Thereafter the cell culture was immediately cooled on ice, and the Cells were grown at 37~ in complete DME (4.5 g/liter glucose) sup- cells were removed and collected by centrifugation at 4"C in a 50-ml tube plemented with 15% FCS (Gibco Laboratories, Karlsrnhe, FRG) in an at- (Falcon Labware). After a wash in ice-cold PBS, the cells were resuspended mosphere containing 10% C(h. For all labeling experiments, cells were at a density of 1 • 107 cells/mi in complete DME (supplemented with collected by centrifugation at 150 g for 5 min, washed once in ice-cold PBS, 0.1% FCS and 2 mM D(+)-galactose) that had been preequilibrated in a 10% and resuspended after centrifugation as described in the sections below. Un- CO2 atmosphere at 370C. Immediately after resuspension, two aliquots less indicated otherwise, labeling was performed in 3.5- or 5-cm petri dishes were removed and cooled on ice (zero time values). The cell suspension re- in a 10% CO~z incubator. For all labeling experiments we used labeling maining in the 50-ml Falcon tube was returned to the incubator and placed DME (sulfate-free and L-tyrosine-free) that was supplemented with 4.5 g/li- on a slowly rotating platform during the chase period. Further aliquots of ter glucose and 0.1-10% FCS dialyzed against 10 mM Hepes, pH 7.3 and the well-suspended cell culture were removed at the indicated times and im- 150 mM NaC1. To increase the efficiency of [35S]sulfate-labeling, a lower mediately placed on ice. Cells and media were further processed as de- concentration of L-methionine and L-cysteine (2% of normal) was used in scribed under Immunoaflinity Purification of IgM. some experiments (Baeuerle and Huttner, 1986). This had no detectable influence on protein synthesis as monitored by the incorporation of [3H]tyrosine into proteins. 20~ Block Cells were resuspended in a 50-mi tube at a density of 1 x 107 cells/ml in Long-term Labeling labeling DME supplemented with 0.1% dialyzed FCS and allowed to equilibrate for 5 h on a slowly rotating platform in the incubator at 37~ The Cells were resuspended at a density of 1 x 107 cells/ml in labeling DME tube was then closed to preserve the atmosphere, removed from the incuba- supplemented with 0.1-10% dialyzed FCS. Cells were kept in the incubator tor, and placed in a 20~ waterbath for 5 rain with occasional mixing. Cells for 30 min before the addition of 18.5 MBq/ml [358]sulfate, 1.85 MBq/ml were then labeled at 20~ for 10 rain with 37 MBq/ml [35S]sulfate in the [3H]tyrosine, 0.11 MBq/ml [14C]tyrosine, 0.93 MBq/rnl [3H]fucose, 1.85 closed tube. Thereafter the tube was placed on ice, the cells were collected MBq/mi [3H]mannose, or 0.93 MBq/ml [3H]galactose, as indicated. In by centfifugation at 4~ washed in ice-cold PBS, split into several aliquots, some experiments labeling was carried out in the absence and presence of and resuspended in 15-ml tubes in chase medium (complete DME sup- 1 pg/ml tunicamycin (Boehringer Mannheim, Mannheim, FRG) or 10..6 M plemented with 0.1% FCS) of either 0, 20, or 370C that had been pre- monensin (Calbiochem-Behring Corp., Frankfurt, FRG), which were equilibrated in all cases in a 10% CO2 atmosphere. The tubes were closed added immediately after resuspension of cells, followed by addition of iso- and placed either on ice (for the zero values) or in waterbaths of 20 or 370C, topes 1 h later. After 18 h of labeling, cells and media were further processed respectively. Aliquots of the cell cultures in the waterbaths were collected as described in Immunoaffinity Purification of IgM. after 15 and 30 min at 20 or 37~ as indicated, and placed on ice. Cells and media were further processed as described in Immunoaffinity Purifica- Short-term Labeling tion of IgM. Cells were resuspended in a 15-mi tube (Falcon Labware, Heidelberg, FRG) at a density of 1 x 107 cells/rnl in labeling DME supplemented with 0.1% CeU-free Sulfation dialyzed FCS that had been preequilibrated in a 10% CO2 atmosphere at 5 x 107 cells were pelleted at 150 g, washed once in ice-cold PBS, and re- 37~ and allowed to equilibrate for 30 min in the incubator. Cells were then suspended in 0.5 ml of sulfation buffer (3 mM imidazole-HCl, pH 7.0, labeled with 37 MBq/ml [35S]sulfate in the tightly closed tube in a 37~ 0.25 M sucrose, 25 mM NaF, 2.5 mM MgCI2, 2.5 mM MnC12, 1 mM waterbath. After the indicated times, aliquots of the cell suspension were 2-mercaptoethanol, 1 mM phenylmethylsulfonylfluoride [PMSF]). The cell removed from the tube and immediately added either to 0.25 vol of a boiling suspension was passed five times through a steel ball homogenizer (Baleh solution of 10% SDS, 50 mM EDTA, or to 0.5 vol of threefold concentrated et al., 1984) with an 8-ttm fitting. The homogenizer was rinsed with 0.5 ml Laemmli sample buffer (Laemmli, 1970) and boiled for 5 rain. The former sulfation buffer and the wash pooled with the homogenate. The lysate was samples were used for immunoaffinity purification of IgM followed by the incubated in the presence of 200 ~g/ml hexokinase (Boehringer Mannheim) ricin-binding assay or neuraminidase digestion, whereas the latter samples and 5 mM I)(+)-glucose for 10 rain at 4~ (Schlossmann et al., 1984). 0.96 were directly subjected to SDS-PAGE. MBq/ml [35S]PAPS was added to the supernatant obtained after centrifugation for 10 min at 800 g and the reaction allowed for 2 h at 30~ The reac- Pulse-Chase Experiments tion was stopped by the addition of 1% SDS followed by boiling. An aliquot of the lysate was directly subjected to SDS-PAGE. IgM was purified from For pulse-labeling with [35S]methionine, washed cells were resuspended at the remaining lysate as described in Immunoaffinity Purification of IgM. a density of 1 x l0s cells/ml in labeling medium (0.1% FCS; no methionine and cysteine) and labeled, after preequilibration for 90 min in the incubator, with 3.7 MBq/ml [35S]methionine for 3 rain. Cells were then cooled ImmunoaJ~inity Purification of lgM on ice, collected by centrifugation, and resuspended in chase medium (com- After labeling, cells were separated from the medium by centrifugation plete DME supplemented with 15 % FCS) that had been preequilibrated in at 150 g for 5 rain at 4~ Cell culture supernatants were centrifuged at 10,000 g for 10 rain to remove debris. Cell pellets were prepaxed for immu- 1. Abbreviations used in this paper: endo H, endo-13-N-acetylglucosamini- noaffinity purification by boiling for 5 rain in 400 Ixl of an SDS-lysis buffer dase H; PAPS, 3'-phosphoadenosine 5'-phospho[35S]sulfate. consisting of 2% SDS, 20 mM Tris-HCl, pH 7.5, 10 mM benzamidine,

The Journal of Cell Biology, Volume 105, 1987 2656 10 mM e-caproic acid, 5 mM EDTA, aprotinin (300 kailikrein inhibitor Determination of the Stoichiometry of units/ml), and 1 mM PMSE The cleared culture media were used without this treatment. To both cell lysates and culture media 1.5 ml of 10% wt/vol ~rosine Sulfation NP-40 was added and the volume was brought to 15 ml in a 15-ml tube with Immunoaffinity-purified [3H]tyrosine- or [~4C]tyrosine-labeled IgM heavy buffer A (0.5 M NaC1, 10 mM Na2HPO4, 2.7 mM KCI, 1.5 mM KH2PO4, chain separated by SDS-PAGE under reducing conditions was analyzed for 5 mM EDTA, pH 7.5). Then 100 gl of a 50% (packed gel per volume) sus- the stoichiometry of tymsine sulfation as described (Baeuerle and Huttner, pension of anti-mouse IgG-agarose (A6531; Sigma Chemical Co.) equili- 1985). brated in buffer A containing 0.1% (wt/vol) sodium azide were added, followed by incubation for 2 h end-over-end at 4~ The agarose beads were collected by centrifugation for 3 rain at 500 g and washed four times with Quantitation of Radioactive Samples 15 ml of ice-cold buffer A. Bound IgM was eluted from the beads by boiling in 1 ml of the SDS-lysis buffer followed by centrifugation. The immu- Gel pieces were dissolved in H202. Samples were analyzed by liquid scin- noaflinity-purified IgM in the supernatants was precipitated, after the addi- tillation counting in Aqualuma (Baker Co., Frankfurt, FRG). Double- tion of 50 gg bovine hemoglobin as carrier, in 80% acetone at -20~ After labeled samples were counted using two separate windows. The spillover centrifugation for I0 min at 10,0130g, the pellets were dissolved either in of the 3H radioactivity into the 35S window was <1%, and the spillover of Laemmli sample buffer for SDS-PAGE (Laemmli, 1970), in O'Farrell lysis the 35S radioactivity into the 3H window ,,.,4%. buffer for two-dimensional PAGE (O'Farrell, 1975), or in dilute NaOH for endo-13-N-acetylglucosaminidase H (endo H) digestion, the ricin-binding assay, and neuraminidase treatment as described below. Results Endo H Digestion The Heavy Chain of lgM is Sulfated Acetone pellets containing immunoaffinity-purified IgM were dissolved in Cell cultures of the hybridoma cell line 81-O4 were labeled 20 gl 0.1 N NaOH and immediately brought to a pH of m6.0 by the addition with [3H]tyrosine or [35S]sulfate, and intracellular and se- of 400 I~1 of a solution containing 100 mM sodium citrate, pH 5.5, 0.1% creted IgM was immunoaffinity-purified by anti-mouse IgG- SDS, aprotinin (300 kallikrein inhibitor units/nil), and 1 mM PMSE Sam- ples were incubated overnight at 300C in the presence or absence of 15 mU/ agarose recognizing the light chains of IgM. After SDS-PAGE ml endo H (Seikagaku, Tokyo, Japan). Proteins were precipitated using 80% under nonreducing conditions (not shown), the secreted acetone, and the resulting pellets were dissolved in Laemmli sample buffer IgM had a slower mobility than the thyroglobulin dimer (Mr for SDS-PAGE. 669,000), indicating that it was present in the pentameric 19S form. Under reducing conditions, two intracellular forms of Ricin-binding Assay the [3H]tyrosine-labeled IgM heavy chain could be distin- Acetone pellets containing immunoaftinity-purified IgM were dissolved in guished with apparent molecular masses of 71 and 76 kD 20 I.tl 0.1 N NaOH and immediately brought to a pH of,o7.5 by the addition (Fig. 1 A, fifth lane). (The proportion of the 71-kD form to of 400 gl buffer B (0.I M Tris-HCl, pH 7.0, 1 M NaC1, 1 mM CaC12, the 76-kD form was found to vary with the age of the cul- 1 mM MgC12). Then 80 gl of a 50% (packed gel per volume) suspension of ricin-agarose (I.2757; Sigma Chemical Co.) equilibrated in buffer B was tures.) In the medium, the heavy chain appeared as a diffuse added. After 2 h of incubation at room temperature on a Vibrax shaker, free band with a molecular mass ranging from 76-86 kD (Fig. and ricin-bound IgM were separated from each other by pelleting of the 1 A, sixth lane). Heavy and light chains of IgM represented agarose beads for 5 min at 500 g. Protein in the supernatant was precipitated the major secretory proteins of the medium (Fig. 1 A, second using 80% acetone and dissolved in Laemmli sample buffer. The pelleted agarose beads were washed twice with buffer B, followed by elution of the lane). When IgM secreted by [35S]sulfate-labeled cells was bound IgM by boiling in Laemmli sample buffer. Samples were subjected immunoaffinity-purified and analyzed on SDS gels under to SDS-PAGE. reducing conditions, the heavy chain was found to contain [35S]sulfate (Fig. 1 A, eighth lane). SDS-PAGE of immuno- Neuraminidase Treatment affinity-purified secreted IgM under nonreducing conditions Acetone pellets containing immunoaffinity-purified IgM were dissolved in indicated that the [35S]sulfate label was contained in the 19S 20 gl of 0.1 N NaOH and immediately brought to pH 5.5 by the addition form of IgM (data not shown). Intracellularly, only the 76- of 400 I~1 of a solution of 1130 mM sodium acetate, pH 5.0, 9 mM CaCI2, kD form of the heavy chain of IgM incorporated [35S]sulfate 150 mM NaCI, aprotinin (300 kallikrein inhibitor units/ml), and 1 mM (Fig. 1 A, seventh lane). No incorporation of [35S]sulfate PMSE Samples were incubated overnight at 30~ in the presence or absence of 0.25 U/ml neuramiuidase from Clostridium perfringens (type X; was observed into the light chains (Fig. 1 A, eighth lane). Sigma Chemical Co.). Proteins were then acetone precipitated for SDS- The IgM heavy chain was a major sulfated protein in the PAGE. medium (Fig. 1 A, fourth lane). In the cells, several sulfated bands in addition to the heavy chain were observed (Fig. 1 Polyacrylamide Gel Electrophoresis A, third lane) but most of them were apparently not secreted SDS-PAGE was performed as described by Laemmli (1970). Processing of into the medium. The sulfation of IgM heavy chain in the hy- gels after electrophoresis was as described previously (Lee and Huttner, bridoma clone 81-04 was not a singular observation. Two 1983). other IgM-secreting cell lines were tested, a rat hybridoma cell line (clone 13/305/9) and the mouse myeloma cell line ~yrosine Sulfate Analysis of the IgM Heavy Chain WEHI 279.1. In both cases the IgM heavy chain was found Immunoaffinity-puriffed [35S]sulfate- or [35S]sutfate/[3H]tyrosine double- to be sulfated (data not shown). labeled IgM heavy chain separated by SDS-PAGE under reducing conditions was analyzed by alkaline hydrolysis for the presence of tyrosine sulfate Sulfate Is Bound to ~rosine and as described previously (Lee and Huttner, 1983; Huttner, 1984). Pronase N-linked Oligosaccharides digestion (Huttner, 1984) of IgM heavy chains in gel pieces was performed using 1 mg/ml pronase (Boehringer Mannheim) in 50 mM ammonium bi- Immunoaffinity-purified [3~S]sulfate-labeled IgM heavy chain carbonate for 22 h at 37~ The digest was tyophilized, redissolved in separated by SDS-PAGE under reducing conditions was sub- 200 gl H20, and excess pronase (as welt as sulfated oligosaccharides) jected to alkaline hydrolysis, a condition in which the tyro- precipitated by 80% acetone. The resulting supernatant was dried in a speed-vac and subjected to two-dimensional thin-layer electrophoresis as sine sulfate ester is stable (Huttner, 1984). Two-dimensional described (Baeuerte and Huttner, 1984; Huttner, 1984). thin-layer electrophoresis of the hydrolysate revealed the

naeuede and Huttner Tyrosine Sulfation in the trans-Golgi 2657 Figure 1. The IgM heavy chain of the hybridoma cell line 81-O4 is sulfated on tyrosine. (A) Hybridoma cell cultures were labeled for 18 h with [3H]tyrosine (3H-Tyr) or [35S]sulfate (35S0~). Aliquots of cell lysates (lanes C) and cell culture media (lanes M) were analyzed, either directly (total) or after immunoprecipita- tion of IgM, by reducing SDS-PAGE. Fluoro- grams are shown. The IgM heavy chains (HC) and light chains (LC) are indicated by brackets. The positions of molecular mass standards are indicated at left. (B) [35S]sulfate-labeled IgM was immunoaffinity purified from the cell culture medium, separated by reducing SDS- PAGE, and the heavy chain was subjected to either alkaline hydrolysis (top) or hydrolysis by pronase (bottom) followed by thin-layer electrophoresis at pH 1.9 in the first dimension (1) and at pH 3.5 in the second dimension (2). Autoradiograms of the thin-layer sheets are shown. The origin is marked by a circle. The positions of tyrosine sulfate (Tyr(S)), serine sulfate (Ser(S)) and threonine sulfate (Thr(S)) standards, as visualized by staining with nin- hydrin, are indicated by dotted lines.

presence of tyrosine sulfate (Fig. 1 B, top). In search for other alkali-labile hydroxyamino acids, [35S]sulfate-labeled IgM heavy chain was subjected to extensive digestion with pronase and analyzed (Fig. 1 B, bottom). Tyrosine sulfate, but no threonine or serine sulfate, was detected. To determine the proportion of tyrosine sulfate relative to the total sulfate incorporated into the IgM heavy chain, cell cultures were double-labeled with [3H]tyrosine and [35S]sulfate. After immunoaflinity purification and SDS-PAGE, radioactivity in the IgM heavy chain was eluted from the gel piece by pronase digestion and the 35SPH ratio was determined in an aliquot of the eluate. Another aliquot of the eluate was subjected to alkaline hydrolysis in barium hydroxide followed by neutralization with dilute sulfuric acid. In this method [asS]sulfate bound to carbohydrates is hydrolyzed and precipitated as barium [35S]sulfate, whereas tyrosine Figure 2. The IgM heavy chain is also sulfated on N-linked carbo- [35S]sulfate remains in the supernatant (Huttner, 1984). Fig. hydrates. Hybridoma cell cultures were double-labeled for 18 h 2 (left) shows that, compared with the eluate (A), the 35S/3H with [3H]tyrosine and [35S]sulfate in the absence (control) and ratio in the neutralized supernatant (B) was decreased by presence of 1 lJg/ml tunicamycin. IgM was immunoaltinity purified 36%. Thin-layer electrophoresis indicated that all of the 35S from the media and separated by reducing SDS-PAGE. Gel pieces in the neutralized supernatant was present as tyrosine sulfate containing the IgM heavy chain were treated with pronase, and the (data not shown). Thus, tyrosine sulfate constituted 64% of ratio of 35S to 3H radioactivity in an aliquot of the eluates was de- the total sulfate present in the IgM heavy chain. termined (columns A). The value found for the heavy chain from To investigate whether the 36% alkali-labile sulfate was control dishes was arbitrarily set to 1.0. The remainder of the elu- bound to N-linked oligosaccharides, cell cultures were la- ates was subjected to alkaline hydrolysis followed by neutralization, and the ratio of 35S to 3H radioactivity in the neutralized superna- beled with [35S]sulfate in the presence of tunicamycin, an tants was determined (columns B). For the control heavy chain the inhibitor of N-glycosylation (for review see Elbein, 1981). ratio was determined in triplicates (bar indicates SD). The recovery Compared with the 76-kD heavy chain of IgM secreted from of 35S radioactivity of tyrosine [35S]sulfate standard in the neutral- control cells (Fig. 2, central panel, left lane), the non- ized supernatant was equal to the recovery of 3H radioactivity of glycosylated heavy chain of IgM secreted from tunicamycin- the [3H]tyrosine-labeled IgM heavy chain (both 90%). (Center) treated cells had a faster mobility in SDS-gels (Mr 61,000) [35S]sulfate-labeled IgM heavy chain of control (left lane) and and was found in lesser amounts in the medium (Fig. 2, cen- tunicamycin-treated (right lane) cell cultures was immunoallinity tral panel, right lane). This 61-kD heavy chain was still sul- purified from the media and subjected to reducing SDS-PAGE. A fated. When the 61-kD IgM heavy chain, double-labeled with fluorogram is shown. The arrowhead on the left indicates the posi- [3~S]sulfate and [3H]tyrosine, was analyzed by alkaline hy- tion of the control heavy chain. The arrowhead with the asterisk on the right indicates the position of the IgM heavy chain secreted by drolysis followed by neutralization, the 35S/3H ratio deter- tunicamycin-treated cell cultures. mined in the pronase eluate (Fig. 2, right panel, A) was al-

The Journal of Cell Biology, Volume 105, 1987 2658 Figure 3. The intracellular 71-kD dimensional cellulose thin-layer electrophoresis/chromatog- _= ~x form of the IgM heavy chain is a raphy (see Baeuerle and Huttner, 1985). The tyrosine sulfate ~ precursor of the 76-kD form. A spot contained 0.72 + 0.07% (255 + 33 cpm, background hybridoma cell culture was pulse- 25 cpm) of the total [3H]tyrosine (35,255 + 3189 cpm). labeled for 3 min with [3~S]methi- (The same percentage was found with [14C]tyrosine-labeled -~. 50 ~._~ onine followedby a 40-rain chase. heavy chain.) This percentage of tyrosine residues present in ~ At the indicated chase times, aliquots of the cell culture (cells sulfated form would equal a stoichiometry of 0.16 mol of plus medium) were lysed in SDS, tyrosine sulfate per mol of heavy chain if the number of tyro- ~" and IgM was immunoaflinity- sine residues in the IgM heavy chain of the 81-O4 cells used 0 s 10! z0 I ~0 purified from the lysates and ana- here is the same as in the mouse IgM heavy chain of clone chase time t min ) lyzed by reducing SDS-PAGE and MOPC 104E (Kehry et al., 1979; 9 tyrosine residues in the fluorography. The radioactivity variable and 13 tyrosine residues in the constant region). in the 71-kD (open circles) and 76-kD (solid circles) forms of the Even if there were no tyrosines in the variable region of the IgM heavy chain was determined and is expressed as percent of the IgM heavy chain of 81-04 cells, the minimal stoichiometry radioactivity in the total IgM heavy chain (71- plus 76-kD form) would be 0.1 mol of tyrosine sulfate per mol of heavy chain present at a given chase time. and thus one mol of tyrosine sulfate per mol of pentameric 19S IgM. In Intact Cells the rER, cis-Golgi, most identical to that in the neutralized supernatant (B). This and Medial Golgi Forms of the IgM Heavy Chain indicated that in the absence of N-linked carbohydrates all Do Not Incorporate Sulfate sulfate incorporated into the heavy chain was present in an The relationship of the two intracellular forms of the IgM alkali-stable form. Thin-layer electrophoresis showed that heavy chain (see Fig. 1, fifth lane) to each other was inves- this alkali-stable sulfate was indeed tyrosine sulfate (data not tigated by pulse-chase labeling with [3SS]methionine (Fig. shown). The finding that tunicamycin treatment prevented 3). At the beginning of the chase period only the 71-kD form the sulfation of IgM in alkali-labile linkage supported the was present. After 40 min of chase ",,50% of the 71-kD form assumption that the 36% of alkali-labile sulfate found in had been converted into the 76-kD form. Thus, the 71-kD the control IgM heavy chain represented sulfate bound to form was a precursor of the 76-kD form. N-linked carbohydrate. Consistent with the presence of sul- The two intracellular forms of the IgM heavy chain were fated N-linked oligosaccharides on the IgM heavy chain was characterized with regard to the state of processing of their the observation that treatment with endoglycosidase F re- N-linked carbohydrates (Fig. 4). The 71-kD form incorpo- moved part of the incorporated radioactive sulfate (data not rated [3H]mannose and trace amounts of [3H]fucose but no shown). Inhibition of N-glycosylation by tunicamycin re- [3H]galactose (Fig. 4, top). Treatment with endo-13-N-ace- duced the relative extent of tyrosine sulfation of the IgM tylglucosaminidase H (endo H), an enzyme known to re- heavy chain by a factor of two (compare columns B in Fig. 2). move N-linked oligosaccharides that have not been modified To determine the stoichiometry of tyrosine sulfation of by medial Golgi-specific enzymes (Tarentino et al., 1974), IgM, the heavy chain of IgM secreted from [3H]tyrosine- markedly increased the electrophoretic mobility of the 71-kD labeled cells was subjected to alkaline hydrolysis, and heavy chain (Fig. 4, lower left). The mobility of the 71-kD [3H]tyrosine separated from [3H]tyrosine sulfate by two- form was unaffected by treatment with neuraminidase (Fig.

Figure 4. Only terminally glycosylateffforms of the IgM heavy chain contain sulfate. (~p) Hybrid- oma cells were labeled for 18 h with [3H]tyrosine (~H-Tyr), [3H]mannose (3H-Man), [3H]fucose (~H- Fuc), [3H]galactose (3H-Gal), or [3~S]sulfate (~SO~). Equal proportions of the IgM immunoaffinity purified from cell lysates (lanes C) and cell culture media (lanes M) were analyzed by reducing SDS-PAGE. Sections of fluorograms are shown. (Bottom) Immunoaffinity-purified[3H]tyrosine-labeled IgM from cell lysates (lanes C) and cell culture media (lanes M) was incubated in the absence (-) and presence (+) of endo H (left) or neuraminidase (right), followed by reducing SDS- PAGE. Sections of fluomgrams are shown. The positions of the 71-kD (.ui) and 76-kD (#) forms of the heavy chain are indicated by open and solid arrowheads, respectively,and by dashed lines. The positions of two molecular mass standards are indicated at right. The open triangle at top (3H-Fuc) indicates the position of a fucose-labeled form of the 71-kD IgM heavy chain.

Baeuerle and Huttner TyrosineSulfation in the trans-Golgi 2659 5, lower right). These results indicated that IgM containing the 71-kD form of the heavy chain was localized in the rER and the cis- and medial Golgi but had not yet acquired the trans-Golgi-specific modifications galactose and sialic acid (Griffiths et al., 1982; Roth and Berger, 1982; Roth et al., 1985; Geuze et al., 1985). (In view of the absence of galactose from the 71-kD heavy chain, we suggest that the small portion of fucose-labeled, 71-kD form [see Fig. 4] was con- rained in IgM that had been processed by a fucosyltransferase localized in the medial Golgi, which acted on the innermost N-acetylglucosamine of the oligosaccharides [Kornfeld and Kornfeld, 1985]. It is unclear whether this fucose-labeled 71-kD heavy chain was still present in medial Golgi cisternae or whether it had already reached the trans-Golgi but not yet acquired galactose.) No detectable amounts of [35S]sulfate were incorporated into the 71-kD form (Fig. 4, top), indicating that neither tyrosine nor carbohydrate sulfation of the IgM heavy chain occurred in the rER, the cis-Golgi, and the medial Golgi before fucosylation. The state of oligosaccharide processing of the intracellular 76-kD form of the IgM heavy chain was very similar to that of the 76-86 kD forms found in the medium (collectively referred to as 76-kD form). In both forms the ratio of [3H]mannose to [3H]tyrosine was less compared with that of the 71-kD form, indicating the loss of mannose residues (Fig. Figure 5. IgM heavy chain short-term labeled with [35S]sulfate in 4, top). Both 76-kD forms incorporated [3H]fucose and vivo contains trans-Golgi-specific modifications. (Top) Hybridoma [3H]galactose. Treatment of the 76-kD forms of cells and cell cultures were labeled for the indicated times with laSS]sulfate medium with neuraminidase increased their mobility in SDS followed by lysis in SDS. IgM immunoattinity purified from cell gels (Fig. 4, lower right), indicating the presence of sialic culture lysates was tested for binding to ricin agarose (lanes P, ricin acid residues on the heavy chain. Thus the intracellular agarose pellet; lanes SN, corresponding supernatant) or incubated 76-kD form of the IgM heavy chain was present in the trans- in the absence (-) and presence (+) of neuraminidase. The IgM Golgi compartment and/or post-Golgi vesicles. (The obser- heavy chain was analyzed by reducing SDS-PAGE. Sections of the fluorogram are shown. The positions of two molecular mass stan- vation that the 76-kD form showed a small increase in elec- dards are indicated at left. (Bottom) Kinetics of pSS]sulfate incor- trophoretic mobility after endo H digestion [Fig. 4, lower poration into the IgM heavy chain. [3SS]sulfatewas added to a hy- left] does not contradict this conclusion, since it has been bridoma cell culture and aliquots of the culture were lysed in SDS reported that two of the five high mannose type oligosaccha- at the indicaWxl times. Samples were directly subjected to reducing rides added to the heavy chain of IgM may remain endo H SDS-PAGE, and the radioactivity in the band corresponding to the sensitive [Hickman et al., 1972].) In contrast to the intracellu- 76-kD form of the IgM heavy chain was determined. The dashed lar 71-kD form, the intracellular 76-kD form did incorporate line represents the extrapolation of the linear increase in [aSS]sul- [35S]sulfate (Fig. 4, top). This indicated that, intracellularly, fate incorporation onto the abscissa. the sulfated form of IgM was localized largely, if not exclu- sively, in the trans-Golgi compartment and/or post-Golgi vesicles. However, the results obtained so far did not exclude came sulfate-labeled during a 4-min pulse bound to ricin- the possibility that sulfation occurred late in the medial Golgi agarose, indicating the presence of galactose (Baenziger and on the 71-kD form of the heavy chain followed by rapid trans- Fiete, 1979), and was sensitive to neuraminidase treatment, port of IgM to the trans-Golgi and conversion of the heavy indicating the presence of sialic acid. When the kinetics of chain to the 76-kD form, which then appeared as the pre- [35S]sulfate incorporation into the IgM heavy chain was in- dominant band upon long-term labeling. vestigated (Fig. 5, bottom) a lag phase of ,02 min was observed before the incorporation started to increase linearly. Most likely this lag phase reflected the time requirements for ~rosine Sulfation of the Heavy Chain radioactive sulfate uptake, PAPS synthesis, and PAPS trans- Takes Place after Exit of lgM from the Medial Golgi location. From the kinetics in Fig. 5 it is apparent that ,080% Hybridoma cells were labeled with [3SS]sulfate for short of the [35S]sulfate found in the IgM heavy chain after label- periods of time and the IgM heavy chain was analyzed for ing for 4 min actually became incorporated during the third the presence of trans-Golgi-specific modifications. The pro- and fourth minute after addition of the label. Thus, during portion of tyrosine sulfate to total sulfate incorporated into incubation of cells with radioactive sulfate for 4 min, IgM the heavy chain after 5 min of labeling was similar to that was effectively labeled only for ,02 min. Since, after this observed after long-term labeling. After immunoaffinity short labeling period, essentially all of the labeled IgM purification of IgM and SDS-PAGE, [35S]sulfate incorpora- heavy chain was in the terminally glycosylated 76-kD form, tion was detectable by fluorography after 4 rain of labeling, it is very unlikely that tyrosine sulfation occurs late in the but not after 1 and 2 min of labeling, and was confined to medial Golgi. the 76-kD form (Fig. 5, top). The 76-kD heavy chain that be- In line with this conclusion were observations made after

The Journal of Cell Biology, Volume 105, 1987 2660 tached to the IgM heavy chain. These findings provided the .w loo -- 100 ... ,s .,- necessary basis for the quantitative comparison of secretion t~ kinetics of galactose- and sulfate-labeled IgM (presented as C a semi-logarithmic plot in Fig. 6). No significant difference in the secretion kinetics of [3H]galactose- and [3SS]sulfate- 5o 50 labeled IgM heavy chains was observed. These secretion kinetics, however, clearly differed from that of [3H]tyrosine- labeled heavy chain, which appeared much later in the t. medium (Fig. 6). These results strongly suggest that sulfate i., ! 1 i i addition to tyrosine (as well as to N-linked carbohydrate) oc- 0 "" 2 5 lo 15 zo 50 loo 1so zso curs at a time point during intracellular transport oflgM very time (rain) similar to that of galactose addition, i.e., in the trans-Golgi, Figure 6. The secretion kinetics of [3H]galactose- and [35S]sulfate- and not much later, e.g., in transport vesicles shortly before labeled IgM are very similar. A hybridoma cell culture was pulse- exocytosis. labeled for 5 min with a mixture of [3H]galactose and [35S]sulfate Plotting the secretion of IgM pulse-labeled with [35S]sul- followed by a 60-min chase. At the indicated chase times, aliquots were removed in triplicate, and ceils and medium separated and fate, [3H]galactose, or [3H]tyrosine on a logarithmic time analyzed directly be reducing SDS-PAGE. The 3H and 35S radio- scale yielded straight lines (linear regression coefficients activity in the band corresponding to the IgM heavy chain was de- t>0.9966) (Fig. 6). Intercepts of the linear secretion kinetics termined by liquid scintillation counting. The labeled IgM heavy with the 50% line gave the t,~ values (t,,~ = 20 min for chain found in the medium at the various chase times is expressed [3H]galactose and [35S]sulfate; t,~ = 94 rain for [3H]tyro- as percent of the total (intraeellular plus secreted) labeled IgM sine). Intercepts of the extrapolated linearized secretion ki- heavy chain ([3H]galactose, solid circles; [35S]sulfate, open cir- netics with the 0% line gave the to values, which are the cles) and is plotted on a logarithmic time scale. Bars indicate SD. times of first appearance of the labels at the cell surface (to In a different experiment, hybridoma cell cultures were pulse- = 2 min for [3H]galactose and [35S]sulfate; to = 26 min for labeled for 10 min with [3H]tyrosine followed by a 4-h chase. At [3H]tyrosine). The secretion of pulse-labeled IgM during the indicated chase times (solid triangles), cells and media were separated, and IgM was immunoaltinity-purifiedfrom cell lysates the course of our measurements can be described as: percent and cell culture media and analyzed by reducing SDS-PAGE. The of total IgM secreted = (50)/(log[t,Jto]) log (t/to). The radioactivity in the [3H]tyrosine-labeledIgM heavy chain was de- first order kinetics observed for the transport of pulse-la- termined and the secretion kinetics is expressed as above. The times beled IgM to the cell surface indicates the existence of a rate- at which the linear secretion kinetics intercepts the 50% line (ha/f- limiting transport step between the sites of sulfation and solid arrows) and, after extrapolation (dashed lines), the 0% line galactosylation (i.e., the trans-Golgi) and the cell surface. (open arrows), are indicated on the abscissa. treatment of cell cultures with the drug monensin (data not Transport of Tyrosine-sulfated lgM to the Cell Surface shown). Monensin has been reported to perturb Golgi pro- is Blocked at 20~ cessing reactions primarily at the level of the trans-cisternae Matlin and Simons (1983) and Saraste and Kuismanen (1984) and to lead to an accumulation of newly synthesized proteins reported that the intracellular transport of viral membrane in the medial Golgi (Griffiths et al., 1983; Tartakoff, 1983). glycoproteins from the Golgi to the cell surface can be ar- After labeling cell cultures with [3H]tyrosine in the pres- rested in a reversible manner by reducing the temperature to ence of 10-6 M monensin, only the 7t-kD form but not the 20~ This led to an accumulation of terminally glycosy- 76-kD form of the IgM heavy chain was found. The 71-kD later viral proteins in the trans-most structure of the Golgi, form accumulated intracellularly and did not incorporate any designated the trans-Golgi network (see Griffiths and Si- [35S]sulfate. mons, 1986). This compartment is further characterized by the presence of galactosyl- and sialyltransferases (Geuze et I~r Sulfation of lgM Occurs al., 1985; Roth et al., 1985). We investigated whether sul- in the Same Compartment as Galactosylation fate-labeled secretory IgM is also reversibly arrested in its To investigate whether sulfation of IgM occurs in the trans- intracellular transport at 20~ When labeling was per- Golgi or later during the intracellular transport in post-Golgi formed at 20~ only the 76-kD form of the heavy chain in- transport vesicles, the kinetics of secretion of IgM pulse- corporated sulfate (Fig. 7, left). The band of cellular heavy labeled with radioactive sulfate was compared with that of chain was broader (76-86 kD) after sulfate-labeling at 20~ IgM pulse-labeled with radioactive gatactose. The amount of than after labeling at 37~ (cf. Fig. 7, left, with Fig. 1 A, and [3H]galactose and [35S]sulfate incorporated into the IgM Fig. 4), which probably reflected a longer residence time heavy chain after a 5-min pulse labeling increased slightly in the trans-Golgi resulting in a higher degree of sialylation during the first 15 min of the chase, with identical rates for and thus slower electrophoretic mobility. The proportion of both isotopes, and plateaued thereafter (data not shown). tyrosine sulfate to the total sulfate incorporated into the This indicated that the specific activity of the cosubstrate for heavy chain was similar (59%) to that observed after long- sulfation (PAPS) and that of the cosubstrate for galactosyla- term labeling at 37~ (64%, see Fig. 2). The incorporation tion (UDP-galactose) decreased rapidly and with similar rel- of radioactive sulfate into IgM at 20~ was reduced by a fac- ative rates during the chase. The observation that the amount tor of 5.15 as compared with 37~ (Ql0 = 2.6). The termi- of [3H]galactose and [3SS]sulfate incorporated into the total nally glycosylated 76-kD IgM heavy chain pulse-labeled for heavy chain (cells plus medium) did not decrease during the 10 min with [35S]sulfate at 20~ was not detectable in the chase showed that both modifications were irreversibly at- medium after 15 min (Fig. 7, left and right) and 30 min (Fig.

Baeuerle and Huttner Tyrosine Sulfation in the trans-Golgi 2661 and carbohydrate sulfation of IgM occur before exit from this compartment.

~rosine Sulfation is the Last Modification of lgM in the trans-Golgi A postnuclear supernatant of an 81-O4 cell homogenate was depleted of endogenous ATP in order to prevent vesicle transport between compartments in vitro (Balch et al., 1984; Davey et al., 1985). The cell-free sulfation of IgM was then investigated using [35S]PAPS under conditions known to allow translocation of this nucleotide into Golgi membrane vesicles (Schwarz et al., 1984). The Golgi vesicles appeared to be largely intact, since the pattern of cell-free protein sulfation was similar to that observed after sulfate labeling of intact cells and was strikingly different from that seen with Figure 7. [3sS]Sulfate-labeledIgM is not transported to the cell detergent-lysed membrane vesicles where tubulin was the surface at 20~ A hybridoma cell culture was pulse-labeled for 10 major sulfated protein (Lee and Huttner, 1985; Huttner and rain at 20~ with [35S]sulfatefollowed by a 30-min chase at the in- Baeuerle, 1987). Only the 76-kD form of the IgM heavy dicated temperatures. At the indicated chase times, cell lysates and chain incorporated sulfate in the cell-free system (Fig. 8), media were analyzed directly by reducing SDS-PAGE followed by and '~50 % of the incorporated sulfate was bound to tyrosine fluorography and determination of the radioactivity in the band cor- (as compared with 64 % after in vivo labeling, see Fig. 2). responding to the 76-kD form of the IgM heavy chain. (Left) The cell-free sulfated IgM heavy chain quantitatively bound Fluorograms showing 15-min chase samples. Brackets indicate the to ricin-agarose, indicating the presence of galactose (Fig. 8, position of the 76-kD (see text) IgM heavy chain (#). Note that the polyacrylamide concentration of the gel with cellular proteins left). After treatment with neuraminidase, the cell-free sul- differs slightly from that of the gel with secreted proteins. (Right) fated IgM heavy chain shifted in its mobility in SDS-gels to Secretion kinetics of IgM (chase at 20~ triangles and solid line; faster-migrating forms, indicating the presence of sialic acid chase at 37~ circles and dashed line). The labeled IgM heavy (Fig. 8, right). These results show that both tyrosylprotein chain found in the medium at the various chase times is expressed sulfotransferase and carbohydrate sulfotransferase are active as percent of the total (intracellular plus secreted) labeled IgM in vesicles that contain the heavy chain of IgM in an already heavy chain. terminally glycosylated form and imply that tyrosine and carbohydrate sulfation of IgM occur after galactosylation and sialylation. If sulfation of the IgM heavy chain occurs after galactosy- 7, right) of chase at 20~ Also, sulfate-labeled macro- lation and sialylation and thus is the last modification before molecules of high molecular mass (presumably proteogly- exit from the trans-Golgi, one may expect that the average cans) were not secreted at 20~ (Fig. 7, left). degree of sulfation of the terminally glycosylated 76-kD The IgM arrested intracellularly at 20~ appeared in the heavy chain from cells is less than that from the medium. medium when the temperature was raised to 37~ (Fig. 7). This was indeed found to be the case, since the ratio of sul- The percentages of sulfate-labeled IgM heavy chain secreted fate/tyrosine incorporation into the cellular 76-kD form was within 15 and 30 min after the temperature was shifted back ~0.75 of that observed for the heavy chain from the medium to 37~ (Fig. 7, right) were very similar to those of galactose- (Table I). labeled IgM secreted in pulse-chase experiments at 37~ (cf. Fig. 6). Even after continuous sulfate-labeling at 20~ for 80 min (at which time ,,o90% of the total sulfate-labeled IgM Discussion was still found intracellularly; data not shown), the propor- We have shown that protein tyrosine sulfation specifically oc- tion of the intracellularly arrested sulfate-labeled IgM that was secreted during a subsequent 20-min chase at 37~ was similar (57%) to that of galactose-labeled IgM secreted in Figure 8. Both tyrosylprotein sul- pulse-chase experiments at 37~ (cf. Fig. 6). This suggested fotransferase and N-linked oligo- (a) that the reduced temperature inhibited a step in the intra- saccharide sulfotransferase activi- cellular transport of IgM that occurred shortly after galac- ties reside in a membrane-enclosed tosylation, and (b) that tyrosine and carbohydrate sulfation compartment containing termi- of the terminally glycosylated 76-kD IgM heavy chain also nally glycosylated IgM. An ATP- took place before this temperature-sensitive transport step. depleted postnuclear supematant In the case of viral membrane proteins in epithelial ceils, the of a hybridoma cell homogenate 20~ block affects the exit from the trans-Golgi (Matlin and was incubated in the presence of [35S]PAPS. Immanoaffinity-puri- Simons, 1983; Saraste and Kuismanen, 1984). Constitutively fled IgM was tested for binding to secreted proteins and viral membrane proteins apparently ricin agarose (P, ricin agarose pellet; SN, corresponding supema- reach the cell surface in the same transport vesicles (Strous rant) or incubated in the absence (-) and presence (+) of neuramin- et al., 1983). In view of these observations it is likely that idase. The IgM heavy chain was analyzed by reducing SDS-PAGE. at 20~ the transport of IgM in hybridoma cells is also Sections of the fluorogram are shown. The positions of two molecu- blocked at the exit from the trans-Golgi, and that tyrosine lar mass standards are indicated at left.

The Journal of Cell Biology, Volume 105, 1987 2662 Table L Relative Sulfate Content of the Cellular and resolved by pulse-chase experiments. It is therefore possible Secreted 76-kD Forms of the IgM Heavy Chain that tyrosine sulfation occurs in both the trans-Golgi cister- 76-kD form Dish [35S]Sulfate [3H]Tyrosine 35S/3Hratio nae and the trans-Golgi network. Under all experimental conditions, the sulfotransferase aC- cpm cpm tivity catalyzing the sulfation of N-linked oligosaccharides Cellular 1 90 598 0.1505 (0.79) on IgM was found to colocalize with tyrosylprotein sul- 2 100 677 0.1477 (0.75) fotransfemse in the trans-Golgi. Moreover, the cell-free sul- Secreted 1 4,423 23,249 0.1902 (1.0) fation of terminally glycosylated IgM in membrane vesicles 2 4,145 20,994 0.1974 (1.0) after addition of [3sS]PAPS implies that the translocation system for PAPS resides, at least in Dart, in trans-Golgi mem- The intraceltular 76-kD form of the IgM heavy chain contains less sulfate than the secreted form. Hybridoma cell cultures were double-labeled with [3H]ty- branes. Thus, the trans-Golgi appears to be the major intra- rosine and [~SS]sulfate for 18 h, and IgM was immunoattinity-purified from cellular site where the various types of protein sulfation the cells and medium followed by reducing SDS-PAGE and determination of Occur. the 3H and 35S radioactivity in the 76-kD forms of the heavy chain. Values are given after subtraction of background, which was 25 cpm for 35S and 19 cpm Tyrosine-sulfated proteins leave the cell by constitutive as for 3H. Numerals in parenthesis give the 3~SPH ratios with the value of the well as regulated pathways (see Huttner and Baeuerle, 1987). secreted 76-kD form arbitrarily set to 1.0. These various secretory pathways are thought to diverge at the level of the trans-Golgi (Kelly, 1985; Griffiths and Si- mons, 1986). The localization of tyrosylprotein sulfotransferase activity in the trans-Golgi is not only consistent with curs in the trans-Golgi. The enzyme catalyzing this post- these observations, but is also economical, since enzymes translational modification (a) was localized via its activity to- with post-trans-Golgi locations would require distinct loca- wards a physiological substrate protein, IgM, rather than via tion signals and retrieval mechanisms. its antigenicity, and (b) was localized in living cells rather It is unclear why an amino acid modification like protein than in subcellular fractions or fixed specimen. Several lines tyrosine sulfation is confined to the trans-Gotgi. In the case of evidence indicate the trans-Golgi localization of the tyro- of N-glycosylation, the localization of specific processing sine sulfation reaction. First, only the trans-Golgi form of enzymes to the endoplasmic reticulum and specific parts of the IgM heavy chain, which could be distinguished from the Golgi can be explained by the sequential formation of the forms localized proximally in the secretory pathway by its final oligosaccharide structures. However, in the case of pro- state of processing of N-linked oligosaccharides, contained tein tyrosine sulfation, the presence of carbohydrates is not tyrosine sulfate in vivo and became tyrosine-sulfated in a requirement for this modification to occur, since unglyco- membrane vesicles in vitro. Second, analysis of the kinetics sylated proteins also become tyrosine-sulfated (see Huttner of secretion indicated that sulfate addition to tyrosine oc- and Baeuerle, 1987). Hence, there is no a priori reason why curred at a time during intracellular transport very similar protein tyrosine sulfation should not occur in a compartment to that of galactose addition to N-linked oligosaccharides. other than the trans-Golgi. One possible explanation is that, Third, IgM arrested in its intraceUular transport in the trans- for reasons of economy, the PAPS-translocating system is Golgi at 20~ incorporated sulfate on tyrosine. Although the confined to a single subcompartment of the Golgi. This may specific trans-Golgi localization of the tyrosine sulfation have evolved to be the trans-Golgi, since PAPS is also the reaction has so far only been shown with IgM as substrate cosubstrate for carbohydrate sulfotransferases, the specific protein, observations made with other secretory proteins, trans-Golgi localization of which may be necessary for the though not conclusive, support the assumption that protein sequential formation of the final oligosaccharide structures. tyrosine sulfation in general is a trans-Golgi-specific event. It is also conceivable that the trans-Golgi localization of Such observations include the inhibition of tyrosine sulfation protein tyrosine sulfation is related to a specific property of of secretogranins I and II by monensin (Lee and Huttner, this compartment. It is worth noting that, within the secre- 1985), and the presence of endo H-resistant oligosaccharide tory pathway, most modifications adding negatively charged on sulfated procollagen V (Fessler et al., 1986). The specific groups to proteins have been shown (sialylation, Roth et al., localization of protein tyrosine sulfation in the trans-Golgi 1985; N-linked otigosaccharide sulfation, tyrosine sulfation, establishes the covalent modification of an amino acid side see this study) or are likely (serine and threonine phosphory- chain as a novel function of this compartment. lation of secretory proteins, Rosa, P., and W. B. Huttner, un- The present data do not allow us to determine with cer- published data) to reside in the trans-Golgi. tainty whether, within the trans-Golgi, protein tyrosine sul- IgM is the first member of the immunoglobulin supergene fation occurs in the trans-Golgi cisternae and/or the trans- family shown to be tyrosine sulfated under physiological Golgi network (Grifliths and Simons, 1986). The results of conditions. Tyrosylprotein sulfotransferase apparently rec- cell-free sulfation of the IgM heavy chain, as well as the com- ognizes tyrosine residues in the vicinity of acidic amino parison of the state of sulfation of the cellular trans-Golgi acids (Lee and Huttner, 1983; 1985) and consensus features form with the form found in the medium, indicated that tyro- of tyrosine sulfation sites have been delineated (Huttner et sine sulfation occurred (at least in part) after galactosylation al., 1986; Hortin et al., 1986; Huttner and Baeuerle, 1987). and sialylation. Hence, tyrosine sulfation appears to be the In the constant part of the mouse IgM heavy chain, the se- last covalent modification of IgM before its exit from the quence surrounding tyrosine 210 (C~tl domain) conforms trans-Golgi. This may suggest a specific localization of best with the consensus features, the only violation being the tyrosylprotein sulfotransferase in the trans-Golgi network. presence of a cysteine residue in position 213, which is However, the lag time between galactosylation and tyrosine disulfide bonded to cysteine 153 (Kehry et al., 1979). sulfation is apparently very short, since it was not clearly Disulfide bonds near tyrosine sulfation sites may be the

Baeuede and Huttner Tyrosine Sulfation in the trans-Golgi 2663 cause of substoichiometric tyrosine sulfation (Huttner and Griffiths, G., R. Brands, B. Burke, D. Louvard, and G. Warren. 1982. Viral membrane proteins acquire galactose in the trans Golgi cisternae during in- Baeuerle, 1987), as observed here for the IgM in which every tracellular transport. J. Cell Biol. 95:781-792. tenth heavy chain was modified. We do not know whether the Griffiths, G., and K. Simons. 1986. The trans Golgi network: sorting at the exit sulfated heavy chains are randomly distributed among the site of the Golgi complex. Science (Wash. DC). 234:438--443. Hickman, S., R. Kornfield, C. K. Osterland, and S. Kornfeld. 1972. The struc- pentameric IgM molecules (one pentamer containing one ture of the glycopeptides of a human gM-immunoglobulin. J. Biol. Chem. sulfated and nine unsulfated heavy chains), whether they are 247:2156-2163. clustered (10% of the pentamers containing only sulfated Hille, A., P. Rosa, and W. B. Huttner. 1984. Tyrosine sulfation: a post- translational modification of proteins destined for secretion? FEBS (Fed. heavy chains and 90% containing only unsulfated heavy Eur. Biochem. Soc.) Lett. 117:129-134. chains), or whether an intermediate situation exists. The Hortin, G, R. Folz, J. I. Gordon, and A. W. Strauss. 1986. Characterization of sites of tyrosine sulfation in proteins and criteria for predicting their occur- functional significance of tyrosine sulfate in the Clxl domain rence. Biochem. Biophys. Res. Commun, 141:326-333. of IgM remains to be elucidated and may be related to the Huttner, W. B. 1982. Sulphation of tyrosine residues-a wide-spread modifica- modulation of an effector function of IgM, the stability of tion of proteins. Nature (Lond.). 299:273-276. Huttner, W. B. 1984. Determination and occurrence of tyrosine O-sulphate in IgM after secretion, or the transcytosis of a subpopulation proteins. Methods EnzymoI. I07:200-223. of IgM molecules. Hutmer, W. B., P. A. Baeuerle, U. B. Benedum, E. Friederich, A. Hille, R. W. H. Lee, P. Rosa, U. Seydel, and C. Suchanek. 1986. Protein sulfa- We thank Professor G. Hartmann and Professor H. Thoenen for support of tion of tyrosine. Colloq. INSERM (Inst. Natl. Sante Rech. Med.). 139: the doctoral thesis ofP. A. Baeuerle; Prof. H. Thoenen for the mouse 81-O4 199-217. and rat 13/305/9 hybridoma lines; Dr. R. Lamers and Dr. G. K6hler for Huttner, W. B., and P. A. Baeuerle. 1987. Protein sulfation on tyrosine. In Modern Cell Biology. Vol. 6. Alan R. Liss Inc., New York. In press. WEHI 279.1 ceils; C. Niehrs for [35S]PAPS; Dis. S. Fuller, K. Simons, Kehry, M., C. Sibley, J. Fuhrmann, J. Schilling, and L. E. Hood. 1979. Amino and G. Warren for stimulating discussions; Dr. B. Dobberstein and Dr, S. acid sequence of a mouse immunoglobulin It-chain. Proc. Natl. Acad. Sci. Tooze for helpful comments on the manuscript; and S. Myers for help with USA. 76:2932-2936. typing. Kelly, R. B. 1985. Pathways of protein secretion in eucaryotes. Science (Wash. DC). 230:25-32. W. B. Huttner was the recipient of grants from the Deutsche Forschungs- Kornfeld, R., and S. Kornfeld. 1985. Assembly of asparagineqinked oligosac- gemeinschaft (Hu 275/3-2, Hu 275/3-3). charides. Annu~ Rev. Biochem. 54:631-664. Laemmli, U. K. 1970. Cleavage of structural proteins during the assembly of Received for publication 27 May 1987, and in revised form 13 August 1987. the head of bacteriophage "I"4. Nature (Lond.). 227:680-685. Lee, R. W. H., and W. B. Huttner. 1983. Tyrosine O-sulfated proteins of PC 12 phaeochromocytoma cells and their sulfation by a tyrosylprotein sulfotrans- References ferase. J. Biol. Chem. 258:11326-11334. Lee, R. W. H., and W. B. Huttner. 1985. (Glu62, Ala3~ TyrO), serves as a Baenziger, J., and D. Fiete. 1979. Structural determinants of Ricinus communis high-affinity substrate for tyrosylprotein sulfotransferase: a Golgi enzyme. aggtutinin and toxin specificity for oligosaccharides. J. Biol. Chem. 254: Proc. NatL Acad. Sci. USA. 82:6143-6147. 9795-9799. Matlin, K. S., and K. Sirnons. 1983. Reduced temperature prevents transfer of Baeuede, P. A., and W. B. Huttner. Inhibition of N-glycosylation induces tyro- a membrane glycoprotein to the cell surface but does not prevent terminal sine sulphation of hybridoma immunoglobulin G. EMBO (Eur. Mol. Biol. glycosylation. CelL 34:233-243. Organ.) J. 3:2209-2215. Roth, J., and E. G. Berger. 1982. Immunocytochemicai localization of galac- Baeuerle, P. A,, and W. B. Huttncr. 1985. Tyrosine sulfation of yolk proteins tosyltransferase in HeLa cells: codistribution with thyamine pyrophospha- 1, 2, and 3 in Drosophila melanogaster. J. Biol. Chem. 260:6434-6439. tase in trans-Golgi cisternae. J. Cell Biol. 93:223-229. Baeuerle, P. A., and W. B. Huttner. 1986. Chlorate-a potent inhibitor of pro- Roth, J., D. J. Taatjes, J. M. Lucocq, J. Weinstein, and J. C. Panlson. 1985. tein sulfation in intact cells. Biophys. Biochem. Res. Commun. 141:870- Demonstration of an extensive trans-mbular network continuous with the 877. Golgi apparatus stack that may function in glycosytation. Cell. 43:287-295. Balch, W. E., W. G. Dunphy, W. A. Braell, andJ. E. Rothman. 1984. Recon- Saraste, J., and E. Kuismanen. 1984. Pre- and post-Golgi vacuoles operate in stitution of the transport of protein between successive compartments of the the transport of Semliki Forest virus membrane glycoproteins to the cell sur- Golgi measured by the coupled incorporation of N-acetylglucosamine. face. Cell. 38:535-549. Cell. 39:405-416. Schlossman, D., S. Schmid, B. Braell, and J. Rothman. 1984. An enzyme that Davey, J., S. M. Hurtley, and G. Warren. 1985. Reconstitution of an endocytic removes clathrin coats: purification of uncoating ATPase. J. Cell Biol. 99: fusion event in a cell-free system. Cell. 43:643-652. 723-733. Dunphy, W. G., and J. E. Rothman. 1985. Compartimental organization of the Schwarz, J. K., J. M. Capasso, and C. B. Hirschberg. 1984. Translocation of Golgi stack. Cell. 42:13-21. adenosine Y-phosphate 5'-phosphosulfate into rat liver Golgi vesicles. J. Biol. Elbein, A. D. 1981. Tunicamycins-usefut tools for studies on glycoproteins. Chem. 259:3554-3559. Trends Biochem. Sci. 6:219-221. Sommer, I., and M. Schachner. 1981. Monoclonal antibodies (O1 to 04) to Farquhar, M. G., and G. Palade. 1981. The Golgi apparatus (complex)- oligodendrocyte cell surfaces: an immunological study in the central nervous (1954-1981)-from artifact to center stage. J. Cell Biol. 91:77s-103s. system. Dev. BioL 83:311-327. Farquhar, M. G. 1985. Progress in unraveling pathways of Golgi traffic. Annu. Strous, G. J. A. M., R. Willemsen, P. van Kerkhof, J. ~: Slot, H. J. Geuze, Rev. Cell Biol. 1:447-488. and H. E Lodish. (1983). Vesicular stomatitis virus glycoprotein, albumin, Pessler, L. I., S. Chapin, S. Brosh, and J. H. Fessler. 1986. Intracellular trans- and transferrin are transported to the cell surface via the same Golgi vesicle. port and tyrosine sulfation of procollagens V. Eur. J. Biochem. 158:511-518. J. Cell Biol. 97:1815-1822. Geuze, H. J., J. W. Slot, G. J. A. M. Strous, K. Hasilik, and K. von Figura. Tarentino, A. L., T. H. Plummer, Jr,, and E Maley. 1974. The release of intact 1985. Possible pathways for lysosomal enzyme delivery. J. Cell Biol. 101: oligosaccharides from specific glycoproleins by endo-b-N acetylglucosamini- 2253-2262. dase H../. Biol. Chem. 249:818-824. Griffiths, G., P. Quinn, and G. Warren. 1983. Dissection of the Golgi complex Tartakoff, A. M. 1982. Simplifying the Golgi complex. Trends Biochem. Sci. I. Monensin inhibits the transport of viral membrane proteins from medial 7:174-176. to trans Golgi cisternae in baby hamster kidney cells infected with Semliki Tartakoff, A. M. 1983. Perturbation of vesicular traffic with the carboxylic iono- forest virus. Z Cell Biol. 96:835-850. phore monensin. Cell. 32:1026--1028.

The Journal of Cell Biology, Volume 105, 1987 2664